JP2004112804A - Fm screening accompanying sub-dot phase modulation - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent introduction of unfavorable artifacts in FM screening without introducing graininess on a halftone image. <P>SOLUTION: A cluster of adjacent pixels can be arranged at any position on a pixel grid in an error diffusion half-toning process. A quantization step of the error diffusion takes an effect of an arrangement of the cluster of pixels into account. Dimensions of the pixel cluster can be fixed or variable. A quantization step of the half-toning process can take an effect of an arrangement of the cluster of pixels into account in the same or different color separation. The method generates less grainy halftone for steadily and uniformly printing in the presence of variation of the printer. A quantization step in the error diffusion process evaluates a quantization value of a possible alternative in arranging a cluster of a pixel. <P>COPYRIGHT: (C)2004,JPO

Description

DETAILED DESCRIPTION OF THE INVENTION

The present invention relates to frequency modulation halftoning, such as frequency modulation performed by error diffusion. The method is particularly suitable for frequency modulated halftoning in systems that print with halftone dots larger than the size of one addressable pixel.

BACKGROUND OF THE INVENTION Screening Objectives A continuous tone image is an image that includes multi-level tone levels without a perceptible quantization step. The purpose of screening (also referred to as halftoning in this document) is to provide a continuous tone image with a printing device having a low tone resolution, such as an offset printing press, which can print only at two tones corresponding to ink and no ink. That is.

The binary halftoning technique converts the density value of a continuous tone image into a geometric distribution of binary halftone dots that can be printed by a binary printing device.

Each halftone dot is represented by one pixel or a clustered set of pixels. A pixel is the smallest addressable element of an image playback device.

If the halftone dots are small enough, the eye will not be able to see the individual halftone dots but will only see the corresponding spatially integrated density values.

２The two main halftone techniques used are known as "amplitude modulation screening" (abbreviated as AM screening), and frequency modulation screening (abbreviated as FM screening).

Amplitude and Frequency Modulation Screen In amplitude modulation screening, the dots that collectively give the impression of a particular tone are arranged on a fixed geometric grid. By changing the size of the halftone dots, images of various gradations can be simulated. FIG. 1 shows the degradation caused by AM screening.

In frequency modulation screening, the distance between halftone dots of a fixed size is modulated to produce various tone values. FIG. 2 shows the same degradation as FIG. 1 resulting from FM screening. Frequency modulation is sometimes referred to as "stochastic screening" because most FM screening algorithms inherently produce stochastic (non-deterministic) dot patterns.

AM and FM halftoning both have advantages and disadvantages.

AM Screening A major advantage of AM halftoning is that it can be efficiently performed by optical, electronic, or digital techniques. A particularly efficient digital implementation is described in U.S. Pat. No. 4,185,304 by Thomas Holladay. A second advantage is, about 7 dots / mm (175 dots / inch, 175dp _i) uniformly and then consistently printed at the image halftoning screen having a commercial offset printing press, produces superior image quality That is what you can do. These two reasons explain why AM screening is the method of choice for most offset printing operations in the graphic arts industry.

In AM halftoning, the grid may be subject to geometric interactions with other patterns because the dots are arranged on a periodic grid characterized by fixed frequencies and angles. These interactions sometimes appear in the form of moire.

The first possible source of moiré is the interaction that occurs between the different halftone dots used in color printing. A solution for handling this form of moiré is presented in US Pat. No. 5,155,599 to Paul @ Delabastita. This patent presents a technique that is convenient and elegant and avoids the introduction of low frequency moire in four-color printing. However, the introduction of rosette-like micromoire typical of four-color printing has not been treated. Even the improvement presented in US Pat. No. 5,828,463 by Paul Delabastita does not remove the rosette-like micro-moire, but only makes it less noticeable. The latter patent also describes that there is a relationship between the relative phase of the screen used in color printing and the resulting color balance. Although the disclosed method improves this color balance over standard screening, it should be noted that this improvement is only effective if the registration between ink separations is perfect. If misregistration occurs, a color balance shift is again introduced.

The second moiré source is the interaction that occurs between the halftone pattern and the addressable pixel grid on which it is located. If the resolution of this grid is too small for the frequency of the halftone screen, a periodic spatial round-off effect occurs that becomes visible as "linen-like" artifacts or as low frequency moiré. This problem can be dealt with to some extent by adding positional noise to the halftone dot centroid. U.S. Patent No. 5,774,229 to Paul @ Delabastita provides such a method. This solution makes it possible to provide a screen with a slightly higher frequency to a recorder with a given resolution. However, when screens with higher frequencies are manufactured, the hemp-like and low-frequency effects recur.

The third moiré source is due to the geometric interaction between the halftone screen and the periodic content in the image itself. This effect is often referred to as "subject moire". Increasing or decreasing the frequency of the halftone screen does not essentially solve this problem, but merely changes the range of frequencies where the problem occurs in the original image. Solutions for treating subject moire are described in U.S. Pat. No. 4,051,536 and Journal @ Opt. {Soc. Amer. {Vol. 66, No. $ 10, $ pp. 985-989, October, 1976, proposed by Paul Roetling in a related paper. His algorithm is known as the ARIES algorithm. Although this method does help to reduce subject moiré, it does not completely eliminate it. Furthermore, it does not address other disadvantages of AM halftoning.

FM Screening FM screening is generally thought to solve the above problems associated with AM screening. Since the center of the halftone dot is not explicitly placed on the fixed grid in the FM screening, the periodic interaction as in the AM screening is not expected.

Three methods are widely used to generate FM screens.

The first method relies on comparing the image to a threshold function on a pixel-by-pixel basis to obtain an FM-screened image. Methods for obtaining such a threshold function are described in U.S. Pat. No. 5,726,772.

Lawrence Ray and James Sullivan describe a second method in WO 91/12686. In this method, a continuous tone image is directly converted to a frequency modulated halftone by addressing a pre-computed bitmap stored in memory in a tone dependent manner.

A third method for frequency modulation was originally invented by Floyd and Steinberg and is called error diffusion. FIG. 3 illustrates how it works. The continuous tone pixel value P has a range from 0.0 (all black) to 1.0 (all white). Modified pixel value P _i of the unscreened image is compared with a fixed threshold value T. If P _i is smaller than T, _{H i} is set to 0.0, a black pixel is printed. Otherwise, H _i is equal to 1.0, a white pixel is printed. Binarization of P _{i introduces} an equal quantization error Ei in P i -H _i. In the error diffusion method, this quantization error value is added to one or more of the unscreened pixels P _{i + x} , _{j + y} . Different pixels receive different fractions of the original error, which is controlled by a "diffusion-weighted" c ₁ to c _n. The sum of the diffusion weights is always one. Since this method works like a feedback loop, the average quantization error value converges to zero in a steady state.

An alternative implementation of error diffusion is shown in FIG. The cumulative diffusion quantization error is kept in a separate buffer and is only added to the original pixel _Pi value just before quantization.

{Robert} Urichney describes a number of improvements over the original error diffusion algorithm in U.S. Pat. No. 4,955,065. The patent describes processing the input pixel values using staggered scanning, adding noise to the threshold, and perturbing the error diffusion weights to obtain a more uniform and isotropic halftone distribution. I have.

顕 著 な Significant improvements in the original error diffusion scheme are also described in US Pat. Nos. 5,045,952 and 5,553,019 by Reiner Eschbach. The disclosures of these patents either modulate the threshold to obtain an edge enhancement effect (first patent) or improve the uniformity of the dot distribution in the high and low intensity regions (second patent). Is one of In US Pat. No. 5,070,413, James @ Sullivan describes an improved screening of color images by performing error diffusion in the colorant vector space, rather than performing scalar error diffusion for each colorant individually. ing. Koen Vande Velde announced a further refinement of this concept at the International Conference on Digital Printing Technology (Proceedings NIP17, IS & T 2001). His algorithm consists of a vector error diffusion scheme, where the quantization of the colors into a set of inks is added in such a way that variations in brightness and corresponding halftoning granularity are minimized in the final output. Is constrained by the output from the preprocessing step. In U.S. Pat. No. 5,565,994, Eschbach proposes a method that aims for a similar purpose but works differently.

改善 A related improvement to our invention is also found in US Pat. No. 5,087,981 to Yee Ng. In this patent, Yee @ Ng describes the use of a printer model to account for dot overlap and compensate for printer gradation non-linearities. U.S. Pat. No. 5,854,882 to Shenge Wang describes a practical method for characterizing printer dot overlap. Similar concepts regarding the introduction of printer models and models of the human visual system are described in Journal of Electronic Imaging, July 1993 Vol. 2 (3), Thrasyvoulos {N. Pappas, Chen-Kong @ Dong, and David @ L. Neuhoff's paper "Measurement of printer printer parameters for model based on halftoning". David @ Neuhoff has patented in U.S. Pat. No. 5,463,472 several of the concepts proposed in this paper. In US Pat. No. 6,266,157, Zhigang @Fan also describes a practical and efficient method of modeling and calibrating the effect of halftone dots in an error diffusion scheme.

Victor @ Ostromoukhov, in his dissertation "A \ Simple \ Efficient \ Error-Diffusion \ Algorithm" published in the bulletin of the SIGGRAPH 2001 conference, adjusts the diffusion weights as a function of the grayscale to produce more uniform nets at various grayscale values. He points out that a point distribution can be obtained.

The standard FM halftoning algorithm implicitly assumes that the size of the printed dot is the same as the pixel size of the addressable grid of the printer, and also matches the pixel size of the original image. This is essentially the situation shown in FIG. This assumption can cause problems if the halftone dots of one pixel size are too small to render properly in the printing process. An example of such a printing process is an electrophotographic printing process. A possible solution to this problem is disclosed in U.S. Pat. No. 5,374,997 to Reiner @ Eschbach, where he describes an error diffusion method that makes the halftone dots n.times.m larger than the size of the addressable pixels of the printer. Advocates its use.

In one of the embodiments shown in FIG. 7, he states that the counter can duplicate the preliminary output pixels of the error diffusion process N times horizontally and M times vertically to obtain larger halftone dots. explained. The output of this scheme when N = M = 2 is shown in FIG.

Although we discussed both AM and FM screening, we need to mention the existence of “hybrid screening”. Delabastita et al. Disclose an example of such a technique in US Pat. No. 5,766,807. According to this technique, halftone density modulation is primarily achieved by changing the size of the dots arranged on a fixed grid, while in the highlight the density modulation is the number of fixed size dots. Is achieved by changing

Disadvantages of the State of the Art, Objectives of the Invention While the following focuses on FM screening based on error diffusion, aspects of our invention are readily applicable to usage in combination with other FM screening techniques.

に も か か わ ら ず Despite its advantages and quality, there are still some problems associated with error diffusion FM screening.

The first of the original error diffusions published by the artifacts Floyd and Steinberg near "reasonable tone values" (1/4, 2/4, 3/4, 1/9, 2/9, 3/9, etc.) The problem is that it does not work well around halftone values and tone values that are multiples of 1/4 and 1/3. At and near these tone values, the standard error diffusion algorithm does not produce a halftone dot distribution with a high phase correlation. That is, the dot distribution tends to be locally organized into a regular self-replicating pattern.

To explain why this problem occurs, let's first focus on the behavior around 50% of the Floyd and Steinberg algorithm. When the error diffusion of Floyd and Steiberg is performed for a tone having a gradation value of exactly 50%, all the halftone dots are arranged in a checkerboard shape. This pattern is in fact the most optimal distribution of dots for this tone, since it minimizes the average distance between the dots and thus also the visibility of the dot pattern. However, for tone values slightly above the 50% tone value, the algorithm introduces extra white pixels everywhere to produce the correct average tone value. These extra white pixels necessarily disturb the phase distribution of the checkerboard pattern. FIG. 8 shows an example in which a gradation value of 128/255 is drawn by standard Floyd and Steinberg error diffusion. These local phase shifts would otherwise disturb the regular pattern and be perceived by the eye as disturbing artifacts. A similar situation occurs for tone values slightly below 50%.

Similar problem exists near 75% gradation value. At exactly 75%, the Floyd and Steinberg error diffusion produces a pattern in which one of the four pixels is black and three of the four pixels are white, and all pixels are 2 × 2. It is arranged in a repeating matrix pattern. Above and below this tone value, this regular pattern is disturbed by the introduction of extra white or black pixels. FIG. 9 shows an example of the 192/255 gradation values drawn by the Floyd and Steinberg error diffusion. Similar behavior is exhibited near the tone value of 25% and near the tone value which is a multiple of 1/9 or 1/16.

{Robert} Urichney already recognizes the above problem, and the method he proposes in US Pat. No. 4,955,065 is effective in reducing the undesirable artifacts described above. However, the use of random elements in his algorithm also introduces graininess into the image. Furthermore, his method does not essentially suppress the artifact, but rather diffuses it.

This view is understood by comparing the halftone shown in FIG. 8 drawn with standard Floyd and Steinberg error diffusion to the halftone shown in FIG. 10 drawn using the improved method by Ulichney.

It is an object of the present invention to avoid introducing undesirable artifacts in FM screening without introducing graininess in the halftone image.

Phase correlated dot locations may introduce low frequency graininess or patterns into color prints.The result of correlated dot positions in a single separation is that for color printing, it is the dot position in different ink separations. This is to indirectly derive the phase correlation. This may introduce low frequency artifacts such as patterns and noise. Furthermore, these artifacts shift or change unpredictably when there is misregistration between decompositions.

This will be explained with an example. Imagine colors printed with cyan and magenta ink separations, both having values of 128/255. The Floyd and Steinberg algorithms produce a dot distribution that looks like FIGS. 11 and 12 for these tonal values.

When these two separations are printed one on top of the other in near perfect registration, as in FIG. 13, they have four possible overlapping combinations of inks: no ink, cyan ink only, magenta ink only, Or, overprinting of cyan and magenta inks occurs. Due to the phase correlation of the halftone dots in the decomposition of the original picture, the overlap combination itself is also correlated. In FIG. 13, this leads to two types of regions. In the first type of region, most of the cyan and magenta halftone dots overlap each other, resulting in a matrix of cyan and magenta overprinted halftone and white spaces. In the second type of region, most of the cyan and magenta halftone dots are located between each other, resulting in a matrix of magenta and cyan halftone dots with no or little white space. When white and cyan are added to the magenta halftone dot, the colors are not exactly the same as when the cyan and magenta halftone dots are added, so the two types of regions have different colors. The end result is that the color balance of the entire print reproduction is not stable and the print appears to be full of stains.

If the register between the two separations changes as in FIG. 14, for example, due to some mechanical instability of the printer or support, the first type of area may change to a second type of area, or the like. Conversely, it may change. Thus, not only is the color balance unstable throughout the print, but it varies with the register of the printer and is unpredictable in the presence of misregistration. As shown in both FIG. 13 and FIG. 14, the correlation artifacts of the individual decompositions are low frequency patterns that were not present in any of the original image decompositions, and that change as a function of the register between the original image decompositions. Can occur.

What the above description shows is that halftone dot location correlation can result in low frequency graininess and pattern formation, which can result in locally unstable color balance when there is misregistration. That is.

Existing techniques use the introduction of random elements, such as weight perturbation in error diffusion or the addition of noise to the threshold, to split the phase-correlated dot positions, but as mentioned earlier, this is halftoning. Graininess is also introduced into the processed image.

It is an object of the present invention to improve the stability and predictability of the color balance in frequency modulation halftoning for color printing and avoid the introduction of low frequency artifacts without introducing graininess in the image.

The fixed dot size of error diffusion represents a compromise The method presented by Eschbach in U.S. Pat. No. 5,374,997 discloses the generation of a dot consisting of adjacent pixels N.times.M times larger than the size of one pixel of a printing device. In practice, the selection of optimal values for "N" and "M" represents a compromise between print stability, graininess, and detail rendition.

A major source of print density and color balance instability is the effect near the dot boundaries (deterministic or stochastic in both space and time). This explains why large halftone dots with short border lengths to cover the same total area are generally more stably printed than small halftone dots. This assertion favors the use of larger halftone dots consisting of clusters of more neighboring pixels. This is especially true for midtone reproduction, since the total amount of halftone dot perimeter and printer variability effects are maximized in the midtone portion of the tone scale.

However, the use of large fixed size halftone dots has disadvantages. For a given tone value, the use of large halftone dots also increases the average distance between them, especially for highlights and shadow tones where this distance is already at its maximum, which is the visibility of the halftone pattern. And may lead to a grainy appearance. Therefore, this assertion actually favors the use of clusters consisting of a small number of pixels.

Large fixed size halftone dots can also impair detail rendition. For example, for proper rendering of a fine-grained texture, printing with small pixel clusters is advantageous. In the case of text and line art rendering, it is even preferable to have a cluster size of only one pixel, since full resolution rendering of the contours of these elements is possible.

These observations point to the need to devise a new frequency modulation halftoning method, the goal of which is to minimize the graininess of halftoning and simultaneously optimize the rendering of textures, fines, text, and line art, while at the same time Optimizing color balance and contrast stability even when there are variations in the printing process.

[Prior art]
Heretofore, the following related documents are known.
Patent Publication 1: U.S. Pat. No. 4,051,536 "Electronic halftone imaging system"
Patent Publication 2: U.S. Pat. No. 4,185,304 "Electronic halftone screening"
Patent publication 3: U.S. Pat. No. 4,955,065 "System for producing dithered images from continued-tone image data"
Patent Publication 4: U.S. Pat. No. 5,045,952 "Method for edge enhanced error diffusion"
Patent Publication 5: U.S. Pat. No. 5,070,413, "Color digital halftoning with vector error diffusion"
Patent Publication 6: U.S. Pat. No. 5,087,981 "Error diffusion of overlapping dots"
Patent Publication 7: U.S. Pat. No. 5,155,599 "Screening system and method for color reproduction in offset printing"
Patent Publication 8: US Patent No. 5,374,997, _{"H i gh addressability error diffusion with minimum} mark size "
Patent publication 9: U.S. Pat. No. 5,463,472 "Model-based halftoning"
Patent Publication 10: U.S. Pat. No. 5,535,019 "Error diffusion halftoning with homogeneous response in high / low intensity image regions"
Patent Publication 11: U.S. Pat. No. 5,553,020 "Void and cluster apparatus and method for generating dither templates"
Patent Publication 12: U.S. Pat. No. 5,543,941 "Method and Apparatus for Halftone Rendering of a Gray Image Using a Blue Noise Mask"
Patent Publication 13: U.S. Pat. No. 5,565,994, "Multiple separation error diffusion, with cross separation correlation control control color images"
Patent Publication 14: U.S. Pat. No. 5,708,518 "Method and Apparatus for Halftone Rendering of a Gray Scale Image Using a Blue Noise Mask"
Patent Publication 15: U.S. Pat. No. 5,726,772, "Method and Apparel for Halftone Rendering of a Gray Scale Image Using a Blue Noise Mask"
Patent Publication 16: U.S. Pat. No. 5,745,259 "Void and cluster apparatus and method for generating dither templates"
Patent Publication 17: U.S. Pat. No. 5,766,807, "Halftone screen and methods for making and using the same"
Patent Publication 18: U.S. Pat. No. 5,774,229 "Halftone screen generator and halftone screen and method for generating same"
Patent Publication 19: U.S. Pat. No. 5,828,463, "Printing plate for a halftone image having tone toner dependent rosette structures BV phase modulation"
Patent Publication 20: U.S. Pat. No. 5,854,882 "Halftone correction systems"
Patent Publication 21: U.S. Pat. No. 5,912,745, "Voil and cluster apparatus and method for generating dither templates"
Patent Publication 22: U.S. Pat. No. 6,172,773, "Voil and cluster apparatus and method for generating dither templates"
Patent Publication 23: U.S. Pat. No. 6,266,157, "Method of error diffusion using 2X2 color correction and increment matching"
Patent publication 24: WO91 / 12686 "Digital halftoning with correlated minimum visual modulation patterns"
Non-patent literature:
Roetling, "Halftone Method with Edge Enhancement and Moire Suppression", JOSA, vol. 66, pp. 985-989, 1976.
-T. N. Pappas, C.I. -K. Dong and D.M. L. Neuhoff, "Measurement of printer parameters for model-based halftoning", Journal of Electronic Imaging, vol. 2, p. 193-204, July, 1993.
-Victor Ostromukhov. A Simple and Efficient Error-Diffusion Algorithm. In Proceedings of SIGGRAPH 2001, ACM Computer Graphics, Annual Conference Series, pp. 147-64. 567-572, 2001.

SUMMARY OF THE INVENTION The above-mentioned advantageous effects are realized by a method having certain features as set forth in claim 1. Particular features of preferred embodiments of the invention are set out in the dependent claims.

Additional advantages and embodiments of the present invention will become apparent from the following description and drawings.

The present invention is based on the observation that frequency modulation halftoning, when taken alone, is not necessarily stochastic in nature. The fixed-size halftone dots are arranged in a periodic rectangular or hexagonal grid, the distance between all the halftone dots is the same, and the modulation of the gradation value continuously changes the distance between the halftone dots. It is desirable to have a frequency modulation technique such as that achieved by: At each gray level, the halftone distribution is completely deterministic and periodic.

It is clear that such frequency modulated halftones are not affected by phase correlation artifacts near reasonable gradation values as described above, while at the same time the level of granularity of the dot distribution is very low. . Thus, such halftoning techniques approximate the characteristics of an ideal frequency halftoning algorithm.

Unfortunately, the modulation of the frequency of the continuous halftone dot distribution can place halftone dots with infinitely high accuracy, thus requiring a recording system with infinitely high addressability. Such a recording system does not exist in the real world.

By using the method of the present invention, the characteristics of the ideal frequency modulation technique described above are not infinite, but use the addressability of the recording system, which is significantly higher than the size of a halftone dot consisting of clusters of adjacent pixels. Is approximated by By approximating the ideal of continuous frequency modulation by allowing the distance between the dots to be modulated in increments much smaller than the size of the dots themselves, as well as suppressing artifacts near reasonable gradation values A frequency modulation halftoning system is provided that offers the advantages of We call this principle "subdot phase modulation." Since the position of the halftone dot is controlled with an accuracy exceeding the size of the halftone dot by the sub-dot phase modulation, the relative position between the halftone dots can be controlled with much higher accuracy than the standard frequency modulation. This principle also opens up the possibility of much better control of color printing artifacts due to the correlated dot positions of the different separations.

The error diffusion method having the specific features of the claims can modulate the distance between halftone dots by taking into account the effect of a change in density value that occurs in a certain area of the output image due to the arrangement of pixel clusters.

It is characterized in that a halftone dot composed of a plurality of adjacent pixels can be arranged at an arbitrary position on a pixel grid.

Having described the general concept of our invention, we want to describe the first embodiment of sub-dot phase modulation.

Having described the basic principles of our invention, the details of a preferred embodiment for performing frequency modulation by subdot phase modulation according to our invention will now be described.
BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 shows the degradation obtained by AM screening.
FIG. 2 shows the same degradation as FIG. 1 obtained by FM screening.
FIG. 3 shows a basic error diffusion method.
FIG. 4 shows an alternative implementation of the error diffusion scheme.
FIG. 5 shows frequency modulation halftoning of a 5/9 color tone without pixel duplication.
FIG. 6 shows frequency modulation halftoning of a 5/9 tone with 2 × 2 pixel duplication.
FIG. 7 shows the results of a method for obtaining clustered halftone dots with improved printing properties using pixel duplication.
FIG. 8 shows the representation of 128/255 tone values obtained by standard Floyd and Steinberg error diffusion.
FIG. 9 shows 192/255 tone values obtained by Floyd and Steinberg error diffusion.
FIG. 10 shows 128/255 tone values obtained by Floyd and Steinberg error diffusion using staggered scanning and threshold 50% noise.
FIG. 11 shows an example of a possible cyan decomposition with a value of 128/255 obtained by Floyd and Steinberg error diffusion.
FIG. 12 shows an example of a possible magenta decomposition with a value of 128/255 obtained by standard Floyd and Steinberg error diffusion.
FIG. 13 shows an overprint representation of the disassembly of FIGS. 11 and 12 in near perfect registration.
FIG. 14 shows an overprint representation of the separations of FIGS. 11 and 12 with the two separations being slightly misregistered.
FIG. 15 shows the result of 2 × 2 pixel duplication following frequency modulation halftoning of a tone having the value 5/9.
FIG. 16 shows the result of 2 × 2 pixel duplication performed after frequency modulation halftoning of a value 5/9 tone in which a random element is introduced in order to eliminate artifacts due to the phase correlation halftone dot position.
FIG. 17 shows frequency modulation halftoning of a 5/9 tone with 2 × 2 pixel duplication and sub-dot phase modulation.
FIG. 18 is a flowchart illustrating an alternative interpretation of the quantization process for standard error diffusion.
FIG. 19 is a flowchart describing a quantization process for error diffusion with sub-dot phase modulation.
20A-G show various examples of quantization by error diffusion with sub-dot phase modulation.
FIG. 21 shows an error diffusion method using sub-dot phase modulation and gradation-dependent variable halftone dot size.
FIG. 22 shows an error diffusion scheme using sub-dot phase modulation and a halftone dot size that depends on both tone value and local contrast.
FIG. 23 shows the rendering of an image with frequency modulation halftoning using state of the art error diffusion.
FIG. 24 shows the result of error diffusion using sub-dot phase modulation.
FIG. 25 shows the results of sub-dot phase modulation and frequency modulation halftoning using variable size halftone dots.

Detailed Description of the Invention Results of Implementation of the Invention FIG. 5 shows an example of a 5/9 value tone obtained by standard error diffusion. 6 and 15 show the same color tone, but in combination with the 2 × 2 pixel replication described in US Pat. No. 5,374,997 to Reiner Eschbach. The tones clearly show artifacts due to the phase correlation halftone dot positions. FIG. 16 again shows the same tone obtained by the same error diffusion, except that the threshold is modulated by random numbers and the weight is perturbed according to the improvement of US Pat. No. 4,955,065 by Robert Urichney. Now, the artifact due to the phase correlation halftone dot position has been improved, but the color tone still looks grainy. FIG. 17 finally shows the tones that are again of the same tone, but this time using a frequency modulation technique using sub-dot phase modulation according to our invention, without introducing random elements. FIG. 17 clearly shows that there is no artifact of the kind shown in FIG. 15, while the graininess is significantly less pronounced than in FIG. Thus, our invention makes it possible to avoid artifacts due to phase-correlated dot positions without introducing halftone graininess.

改善 The improvement shown in FIG. 17 is obtained by the deformation error diffusion process.

Further Definitions The error diffusion process is as follows: an image consisting of raw input pixels with n1 possible values representing the optical density is processed into a processed image with n2 possible values representing the optical density. Defined as a method for converting to an output image represented by pixels, where 1 <n2 <n1. Processing a pixel includes the following steps.
A quantization step of determining the value of the processed pixel from the value of the modified pixel; an error value calculating step of determining a difference between the modified pixel and the processed pixel; at least one of the unprocessed pixels being quantized. Error diffusion step deformed by part

A variant of the error diffusion scheme according to our invention comprises the following steps.
A quantization step of determining the value of at least two processed pixels forming a cluster from the value of the deformed pixel; an error calculating step of determining a difference between the deformed pixel and the processed pixel; Error diffusion step deformed by at least part of the quantization error value

By determining a group of n × m processed pixels during the quantization step, halftone dots that are n × m times larger than the size of each addressable pixel can be obtained. However, since error diffusion is still performed at the full resolution of the recording device, the phase (or position) of the halftone dot is controlled at the full resolution of the recording device. The end result is an error diffusion scheme with subdot phase control that provides the advantages shown in FIG.

The method according to the invention is described in European patent application EP 1 239 662 "improved @ halftoning @ for @ printing @ with @ multiple @ inks" and U.S. Pat. No. 5,535,019 "error @ diffussion @ halftoning @ electronics.com /. Can be used.

The output device reproduces an image by arranging black dots on white paper, for example. The smallest recording element of a recording device is called a pixel. Recording devices typically can only place pixels at specific locations determined by the resolution or addressability of the recording system. This is determined, for example, by the resolution of the recording head or, for example, the minimum steps that a paper transport system can take. All pixels are located on a pixel grid with a particular resolution. The horizontal and vertical resolution can be the same. For example, an electrophotographic device can have a resolution of 16 × 16 pixels / mm (400 × 400 dots / inch). Other systems can have different resolutions in two directions, for example, an inkjet printing system has 192 x 48 dots / mm (4800 x 1200 dpi). The pixel that is the smallest recording element is thus arranged on the pixel grid. A halftone dot is created from a plurality of adjacent pixels. It is the halftoning algorithm used that determines where the dots are located.

Standard Error Diffusion Before describing the method of the present invention in detail, one begins by presenting the standard error diffusion from a somewhat different angle in the literature.

In the following, it is assumed that the origin (0, 0) of the image corresponds to the upper left corner, the value 0.0 represents "black", and the value 1.0 represents "white". Placing halftone dots means printing black dots on a white canvas.

We have modified P (x, y) by the input pixels or input pixel values of the original image and P ′ (x, y) by the error contributions made to the pre-processed pixels that obtained the modified pixel values. A modified pixel of the original image, H (x, y), is defined as a pixel of the output image that has been subjected to the halftoning process. The error diffusion process converts all input pixels P (x, y) to halftone output pixels H (x, y).

す る Before starting the processing, the output pixel values of the halftone processed image are all preset to 0.5. In this approach, the error diffusion process replaces the preset value of H (x, y) with either 0.0 or 1.0 in such a way that an optimal halftone rendering of the original image is obtained. Consisting of

Image processing is usually performed line by line from the top of the image. Within one row, processing may be left-to-right, or right-to-left, or preferably, alternating between these two directions, as suggested by Robert @ Urichney in US Pat. No. 4,955,065. Done. The latter approach is called "staggered scanning."

The processing of pixel P '(i, j) at position (i, j) comprises the following steps, also shown in FIG.
1) Determine a quantization set consisting of two possible quantization values Q ₁ (i, j) and Q ₂ (i, j) corresponding to the next available different case.
a. The gray pixel H (i, j) is converted to a black pixel by reducing the preset halftone value H (i, j) from 0.5 to 0.0. The quantization value Q ₁ (i, j) in this case is 0.0.
b. By increasing the value from 0.5 to 1.0, the gray pixels H (i, j) are converted to white pixels. The quantization value Q ₂ (i, j) in this case is 1.0.
2) If Q ₁ (i, j) is closer to P ′ (i, j), H (i, j) is set to 0.0 and Q (i, j) = Q ₁ (i, j) Otherwise, H (i, j) is set to 1.0 and Q (i, j) = Q ₂ (i, j). Using this criterion, a selection of a quantization value from a quantization set is made based on the transformed pixel value.
3) Calculate an error value E (i, j) as a difference between P '(i, j) -Q (i, j).
4) One or more of the raw input pixels P (i, j) of the original image are summed up to a total of 1.0 using, for example, "Floyd and Steinberg" error diffusion weighting. It is deformed by adding a fraction of the error E (i, j) to the value.

Error diffusion with subdot phase modulation
General Embodiment A first possible embodiment of the invention will now be described using the previous description of standard error diffusion as a basis.

Just as in the case of standard error diffusion, processing an image consists of processing the image line by line and pixel by pixel. The order of processing of the pixels in a row can follow a left-to-right, right-to-left, or preferably staggered scan.

One salient feature of the new method is that during the "quantization step" at location (i, j), not only the value of pixel H (i, j) but also the neighborhood around location (i, j) That is, the value of the cluster of the pixel H (x, y) can be changed. Such clusters can be composed of a square, rectangular, or circular array of pixels, or can have any shape. A cluster of adjacent pixels can have a dark value representing "ink spot" or a bright value representing "ink missing". A cluster can include a fixed or variable number of pixels (which ultimately includes one pixel), but at least one of them must include two or more pixels. In the method according to the present invention, taking into account the change in the density value of the area corresponding to two or more pixels of the output image and the overlap of the cluster with the cluster arranged in the previous image processing step, two options ( That is, a quantization set including a quantization value of a dark pixel or a bright pixel cluster arrangement is evaluated. One selected quantization that most closely approximates the modified pixel value P ′ (i, j) is selected, an appropriate quantization error is calculated in consideration of pixel overlap, and this error value is converted into at least one other quantization value. Distributed to input pixels.

The decision process of choosing one of the two alternatives (ie, the arrangement of the dark or bright pixel clusters) involves not only the change in the density value of the output image in the area that coincides with the pixels of the adjacent pixel arrangement cluster, but also possibly its arrangement. It is especially noted that the surrounding area of the cluster affected by is also taken into account.

Using the notation, symbols and conventions in the previous description of standard error diffusion, the present invention can now be generally described by the following sequence of operations, also summarized in FIG.

The processing of the halftone dot at the input pixel position P '(i, j) includes the following steps.
1) Determine two possible quantization values Q ₁ (i, j) and Q ₂ (i, j) of the quantization set corresponding to the next distinguishable available case.
a. A cluster of black pixels is arranged. The quantized value Q ₁ (i, j) in that case corresponds to 0.5−r * 0.5−s * 1.0, where “r” is converted from 0.5 to 0.0 Corresponding to the number of positions, "s" corresponds to the number of pixels that change from white (1.0) to 0.0 by placing this cluster of black pixels. The pixels that went black in the previous step can be turned black again, but such pixels are used in the determination of “r” to avoid the “double counting” of the gray-to-black pixel conversion. Not counted.
b. A cluster of white pixels is arranged. The quantized value Q ₂ (i, j) in that case corresponds to 0.5 + t * 0.5 + u * 1.0, where “t” is the number of positions converted from 0.5 to 1.0. Correspondingly, “u” corresponds to the number of positions that change from black (0.0) to white (1.0). Pixels that were white in the previous step can be whitened again, but such pixels are not included in the determination of “t” to avoid the “double counting” of gray-to-white pixel conversions. Not counted.

Changes in the density values of the regions of the output image in two distinct cases are taken into account.
2) Select a quantization value. If Q ₁ (i, j) is closer to P ′ (i, j), the cluster of output pixels H (i + x, j + y) centered at position (i, j) is set to 0.0 (black), Q (i, j) = Q ₁ (i, j). Otherwise, the cluster of output pixels H (i + x, j + y) is set to 1.0 (white), and Q (i, j) = Q ₂ (i, j).
3) Calculate an error value E (i, j) as a difference between P '(i, j) -Q (i, j).
4) One or more of the unprocessed input pixels P '(i, j) of the original image are added to the values of the error E (i, j) so that the sum of the fractions becomes 1.0. Deform by adding.

Now that the algorithm has been detailed, it is possible to explain why and how the objects of the present invention are achieved.

The first observation is that a new method of error diffusion allows the generation of halftone dots that are substantially larger than one pixel of the addressable grid of the drawing device, and this enlargement improves the printability of the frequency-modulated halftone. That is to achieve.

It is also clear that the location of the dots in the new method is controlled by increments that are substantially smaller than the size of the dots, which makes it possible to place the dots anywhere on the pixel grid. Improved control over the phase of the halftone dot would achieve our goal of suppressing spatial rounding effects otherwise causing disturbing artifacts near reasonable tone values, without introducing noise into the algorithm. ,enable.

Since subdot phase modulation makes more positions available to the dots within a single separation, more control is also needed to avoid the problems associated with the relative positions of the different separation dots in color printing. Will be possible.

で も Even when the dot positions tend to be phase-correlated, this correlation can be eliminated by introducing a smaller amount of randomness into the algorithm than without sub-dot phase modulation.

These observations indicate that for color printing applications with frequency-modulated halftoning, the new method improves the stability and predictability of the color balance and introduces low-frequency artifacts without introducing an unpleasant amount of graininess into the image. We support the statement of avoidance.

However, the placement of white and black clusters is constrained by additional rules. This means that not all cases of placing a cluster are possible or available. Not all quantized values are calculated.

First specific embodiment: "white" clusters cannot be overprinted with "black" clusters In a first exemplary and representative specific embodiment, in a given processing step the pixels of the cluster are Constraints are placed on the rules for how the pixels of a placed cluster can be changed. For example, it can be stated that a previously white pixel can change its color to black in the next processing step, but not vice versa. This rule states that once ink is placed on white paper, it can no longer be undone, but it is completely possible to print ink spots that partially overlap previously printed ink spots. Corresponds to intuitive concepts.

The processing of the halftone dot at the pixel position P ′ (i, j) in this embodiment includes the following steps.
1) Determine a quantization set of two possible quantization values Q ₁ (i, j) and Q ₂ (i, j) corresponding to the next distinguishable case.
a. A cluster of black pixels is arranged. The quantized value Q ₁ (i, j) in that case corresponds to 0.5−r * 0.5−s * 1.0, where “r” is converted from 0.5 to 0.0 Corresponding to the number of positions, "s" corresponds to the number of pixels that change from white (1.0) to 0.0 by placing this cluster of black pixels. The pixels that went black in the previous step can be turned black again, but such pixels are used in the determination of “r” to avoid the “double counting” of the gray-to-black pixel conversion. Not counted.
b. A cluster of white pixels is arranged. The quantized value Q ₂ (i, j) in that case corresponds to 0.5 + t * 0.5, where “t” corresponds to the number of positions converted from 0.5 to 1.0. This counting scheme reflects the assumption that black pixel H (x, y) = 0.0 cannot change again to white H (x, y) = 1.0.
2) Select a quantization value. If Q ₁ (i, j) is closer to P ′ (i, j), the cluster of output pixels H (i + x, j + y) centered at position (i, j) is set to 0.0 (black), Q (i, j) = Q ₁ (i, j). Otherwise, the cluster of pixel H (i + x, j + y) is set to 1.0 (white), and Q (i, j) = Q ₂ (i, j).
3) Calculate an error value E (i, j) as a difference between P '(i, j) -Q (i, j).
4) Add a fraction of the error value E (i, j) to one or more of the raw input pixels P (i, j) of the original image so that the fractions add up to 1.0. To obtain a modified pixel value.

FIG. 21 shows a number of representative cases that we want to use to demonstrate the principle of quantization in the above embodiment. Not all possible cases are shown, and what is shown does not necessarily correspond to the same tone values as the original image. It will be clear that alternative cases and situations are easily derived using the same principles.

も Both the black and white halftone dots in FIG. 21 are assumed to be 2 × 2 pixel clusters. Processing is performed from left to right. The position (i, j) in the drawing corresponds to the upper left position of the 2 × 2 halftone cluster. The numbers in the squares represent the values H (x, y) before the quantization step. The first quantization option Q1 (i, j) always corresponds to the arrangement of black dots, and the second quantization option Q2 (i, j) corresponds to the arrangement of white dots. Assume that black pixels can never be white again. As described above, the purpose of calculating Q1 (i, j) and Q2 (i, j) is to determine which of the two values most closely approximates the modified pixel value P ′ (i, j). A determination, after which the value of H (x, y) is finally determined, a quantization error is calculated and diffused to one or more unprocessed pixels. Consider now the various available cases of FIGS. 20A-20G.

Case 1: The values of H (x, y) before the quantization step are all 0.5. Q1 (i, j) is 0.5-4 * 0.5 = -1.5 because this option reduces the value of four gray pixels from gray to black. The quantized value of Q2 (i, j) is 0.5 + 4 * 0.5 = 2.5 because in this option the values of the four gray pixels increase from gray to white.

Case 2: Two of the values of H (x, y) before quantization are 1.0, and the other two are 0.5. Q1 (i, j) is 0.5-2 * 0.5- because this option reduces the values of the two gray pixels from gray to black and also reduces the values of the two white pixels to black. 2 * 1.0 = -2.5. The quantized value of Q2 (i, j) does not change because the value of two pixels increases from gray to white and the value of the other two pixels is already 1.0, so that the value of. 5 + 2 * 0.5 = 1.5. Their increase from gray to white has already been taken into account in the previous processing step, and no further consideration is needed in this step.

Case 3: Two of the values of H (x, y) before quantization are 0.0, and the other two are 0.5. Q1 (i, j) indicates that the values of the two gray pixels have been reduced from gray to black, the values of the two black pixels have not changed, and their value has decreased from 0.5 to 0.0. 0.5-2 * 0.5 = -0.5. The quantized value of Q2 (i, j) is 0.5 + 2 * 0 because two pixels are converted from gray to white and two black pixels cannot be overwritten on white pixels and their values do not change. .5 = 1.5. The reduction of these values from 0.5 to 0.0 has already been taken into account in the previous processing step and need not be considered any more in this step.

Case 4: One of the values of H (x, y) before quantization is 1.0, and the other three are 0.5. Q1 (i, j) is 0.5-3 * 0.5-1 * because this option reduces the value of three gray pixels to black and the value of one white pixel to black. 1.0 = −2.0. The quantized value of Q2 (i, j) is 0.5 + 3 * 0.5 = 0.5 + 3 * 0.5 since the value of three pixels increases from gray to white and the value of one pixel that was already white does not change. 2.0. Its increase from 0.5 to 1.0 was taken into account in the previous processing step, which need not be considered anymore in this step.

Case 5: Two of the values of H (x, y) before quantization are 0.0, one is 0.5, and the fourth value is 1.0. Q1 (i, j) is 0.5-1 * 0.5-1 * because this option reduces the value of one gray pixel to black and the value of one white pixel to black. 1.0 = −1.0. The other two pixels do not change because they are already black, and their reduction from 0.5 to 0.0 has been taken into account in a previous processing step. The quantization value of Q2 (i, j) is 0.5 + 1 * 0.5 = 1.0. Since the two black pixels cannot overwrite the white pixels, their values are preserved and the reduction of their values from gray to black has already been taken into account in the previous processing step and is taken into account further in this step No need to do. One pixel is converted from gray to white and another remains unchanged because it is already white and no longer contributes to Q2 (i, j) in this processing step.

Case 6: One of the values of H (x, y) before quantization is 1.0, one is 0.0, and the other two are 0.5. Q1 (i, j) is determined by this option because the value of two gray pixels is reduced to black, the value of one white pixel is reduced to black, and the value of the other black pixel remains unchanged. , 0.5-2 * 0.5-1 * 1.0 = -1.5. The quantization value of Q2 (i, j) is 0.5 + 2 * 0.5 = 1.5 because the value of two pixels increases from gray to white and the value of the other two pixels does not change. It is.

Case 7: Two of the values of H (x, y) before quantization are 1.0, one is 0.0, and the other is 0.5. Q1 (i, j) is because this option reduces the value of one gray pixel to black, the value of two white pixels to black, and the value of the other black pixel does not change. , 0.5-1 * 0.5-2 * 1.0 = -2.0. The quantized value of Q2 (i, j) is that the value of only one pixel has increased from gray to white and the other three pixel values have already been white or they have been previously black. 0.5 + 1 * 0.5 = 1.0, since there is no change because it cannot return to white.

Second specific embodiment: In another exemplary and representative embodiment where the "white" cluster has a size of one pixel , the black clusters have a fixed rectangular size and overlap each other or white space. While printing is possible, white "clusters" consist of only one pixel and cannot be overprinted on black pixels.

Assume that the size of the black halftone dot is n × m pixels.

If the processing is performed from left to right, the halftone dot is referenced by the upper left pixel (i, j), and if the halftone dot is placed at the position (i, j), 0 ≦ x <m and 0 ≦ y <n The pixel value H (i + x, j + y) becomes 0.0 (black).

When the processing is performed from right to left, the halftone dot is referred to by the upper right pixel (i, j). , The output pixel value H (i + x, j + y) becomes 0.0 (black).

In that case, the processing of the deformed pixel P ′ (i, j) includes the following steps.
1) Determine a quantization set of two possible quantization values Q ₁ (i, j) and Q ₂ (i, j) corresponding to the next distinguishable case.
a. Place black dots. The quantized value Q ₁ (i, j) in that case corresponds to 0.5−f * 0.5, where f is the position converted from 0.5 to 0.0 by arranging halftone dots Corresponding to the number of Positions already at the value of 0.0 as a result of the previous processing step are used to determine Q ₁ (i, j) in order to avoid the “double counting” of the gray-to-black pixel conversion. Not counted.
b. Place a white dot. The following two cases are recognized based on the value of H (i, j). i. H (i, j) was still the preset value of 0.5. In that case, H (i, j) is converted from 0.5 to 1.0 to a white (1.0) pixel, and Q ₂ (i, j) is the gray to white conversion of that pixel. 1.0 (white) is reflected.
ii. Alternatively, H (i, j) has already been set to 0.0 as a result of the arrangement of black dots in the previous processing step. Since a black pixel cannot be a white pixel again, the value of H (i, j) remains 0.0 and Q ₂ (i, j) is H (i, j) at position (i, j). The value of j) is set to 0.5, reflecting that the value of the pixel at the position (i, j) has not changed in the processing.
2) When Q ₁ (i, j) is closer to P ′ (i, j), all the pixels H (i + x, j + y) of the halftone dot at the position (i, j) are set to 0.0, and Q (i , J) = Q ₁ (i, j); otherwise, Q (i, j) = Q ₂ (i, j), and H (i, j) is the case (i) described for case 1b. Alternatively, two values corresponding to (ii) can be taken.
3) Calculate an error value E (i, j) as a difference between P '(i, j) -Q (i, j).
4) One or more of the unprocessed input pixels P '(i, j) of the original image are added to the values of the error E (i, j) so that the sum of the fractions becomes 1.0. Deform by adding.

で Obviously, this method allows for precise control of the location of halftone dots consisting of multiple pixels on the pixel grid. The halftone dots can be located anywhere on the pixel grid to best represent the input image to be transformed. The change in the density value of the area of the output image is considered.

Having described a number of basic embodiments of the method according to the invention, we now turn to a more advanced concept that uses similar principles but additionally provides other enhancements that further improve image quality. to continue.

Third Specific Embodiment: Variable Dot Size In consideration of existing technology, it has already been mentioned that the selection of an appropriate dot size represents a compromise. The graininess of highlight and shadow areas is reduced by printing with smaller dots, but the instability of contrast and color balance benefits from using larger dots. It has also been shown that solid text and graphics contours can be obtained with just one pixel dot size.

In the following, variable dot sizes are used for three different grayscale ranges, Range1, Range2, and Range3, which are separated by a second bordertone value (SecondBorderToneValue) and a first bordertone value (FirstBorderToneValue). Presents a variant of the sub-dot phase modulation scheme, which enables Therefore, the three gradation ranges are as follows.
-Range 1: [0.0, first boundary tone value]
-Range 2: [first boundary gradation value, second boundary gradation value]
Range 3: [second boundary gradation value, 1.0]

用語 Depending on the position of the boundary value of the gradation range, the terms “Shadow boundary” and “Highlight boundary” can be used for the boundary tone value (BorderToneValue). More generally, the cluster size can be adjusted according to the input pixel value.

Assume that the size of the halftone dot can vary between two sizes of n × m pixels or q × p, where q> n and p> m. Hereinafter, the halftone dot size of h × w pixels is represented as h × w, where n ≦ h ≦ q and m ≦ w ≦ p.

If the processing is performed from left to right, assume that the halftone dot is referenced by the upper left pixel (i, j). This means that placing a halftone dot at the position (i, j) affects the pixel value H (i + x, j + y) as 0 ≦ x <w and 0 ≦ y <h. If the processing is performed from right to left, the halftone dot is referenced by the upper right pixel (i, j). Placing a halftone dot at the position (i, j) affects the pixel value H (i + x, j + y) as -w <x≤0 and 0≤y <h.

Using the notation, symbols and conventions in the preceding description, the third embodiment of the invention can now be described in the following order of operation.

The processing of the halftone dot at the pixel position (i, j) comprises the following steps.
1. The halftone dot size is determined as a function of the unmodified pixel value P (i, j).
a. If (P (i, j) <first boundary tone value), (dot size (h, w) is h = q and w = p)
b. If (second boundary gradation value> = P (i, j)> = first boundary gradation value), then (the halftone dot size is the second boundary gradation value from h = q and w = p in the first boundary gradation value). (It changes in proportion to h = n and w = m at the boundary gradation value.)
c. If (P (i, j)> second boundary tone value), (dot size (h, w) is h = n and w = m)
2. The two possible quantized values Q ₁ (i, j) and Q ₂ (i, j) corresponding to the case where the halftone dots can be distinguished or not are determined, the error is calculated, and the unprocessed pixels are determined. The step of deforming according to the amount of error is exactly the same as described above when there is no variable halftone dot size.

Although the above example has been described for the case of a tone range divided into three sub-ranges, it will be apparent to those skilled in the art that the same principles can be used to divide the tone range into any number of sub-ranges. Will. In yet another embodiment, the transition from one dot size to another does not occur at a fixed tone value, but rather spreads over a range of tone values. This effect can be obtained, for example, by determining the dot size based on the sum of P (i, j) and a small random number. With the addition of random numbers, the transition from one dot size to another occurs randomly at slightly higher or lower grayscale values, and the transition from one dot size to another size is A desired effect of diffusing the entire gradation value is obtained.

It is also possible to include a dot distribution changing step in the low or high intensity image area. The above embodiment can also be used to draw solid text and line art at one pixel cluster size so that their outlines are drawn at full resolution. This can be done by setting the pixel output value to the corresponding minimum or maximum output value if the input value is the minimum or maximum possible input value.

Yet another variation of the above embodiment is a network of fewer pixels than for an image region containing a degree of local contrast, for rendering image regions containing a high degree of local contrast, such as textures or subject boundaries. Use points.

To cause this, the undeformed pixels are classified as belonging to regions containing low local contrast, medium local contrast, or high local contrast. Depending on which category the pixel belongs to, a large, medium or small dot size is generated. The classification of the amount of local contrast is based on the measurement of the amount of change in the undeformed pixel value in the area around the undeformed pixel P (i, j). The simplified method simply uses the difference between the minimum and maximum undeformed pixel values of the area around the undeformed pixel P (i, j). A more sophisticated method that relies on the analysis of local neighborhood histograms can quantify the degree of local contrast and control the local dot size.

Fourth specific embodiment: Multi-level error diffusion with sub-dot phase modulation Multi-level frequency modulation is defined as a process by which a halftoned pixel can have more than two quantization values. Certain reproduction processes, such as displays and certain inkjet printing processes, can reproduce multiple density levels for each individual colorant. In inkjet printing, this ability depends on modulation of the size of the droplets, choice of inks with different inks, or a combination of both.

ハ ー フ The halftoning method according to the present invention is also effective for a multilevel halftoning process. For N-level (as opposed to two-level) multi-level frequency modulation, the number of quantization options defined in any of the previous embodiments is simply two (whether to place clusters of black pixels or not). To N.

This will be explained with a representative example. Imagine that the possible quantization values correspond to one set (0.00, 0.25, 0.50, 0.75, and 1.00).

For each pixel cluster, a quantization value must be determined, which value is then assigned to each pixel. This assignment changes the density values of these pixels. Assuming that a given pixel is still at its initial value of 0.5, the next density value change is possible for the multi-level halftoning process.
−0.5 to 0 = −0.5
−0.5 to 0.25 = −0.25
-No change = 0
−0.5 to 0.75 = + 0.25
−0.5 to 1 = + 0.5

全 The total change in density corresponding to all pixels in the halftone dot is easily calculated and serves as a basis for calculating the error diffused to the unprocessed pixels.

Just as in the case of binary halftoning, certain rules and constraints can be applied on how to change the value of a pixel as a function of its previous value. For example, Koen Vande Velde describes a monochromatic method with restrictions on the possible output values in the European patent application EP 1 239 660 "adequate quantification in multilevel halftoning". Constraints are applied to avoid significant deviations in halftone dots by limiting the possible difference between the input value of the pixel and the density of the multi-level output pixel.

制約 Also, restrictions in the case of multilevel color printing are known.

One such method is the method of Koen Vande Velde in European patent application EP 1 239 662 "improved color halftoning for printing with multiple inks". In this way, certain combinations of multi-level color values of output pixels that result in deviating luminance values are excluded from the possible combinations.

It is also clear that the method according to the invention can be applied to multi-level systems with a different number of levels for each color.

Fifth Specific Embodiment: Vector Error Diffusion with Sub-Dot Phase Modulation The above embodiments describe the use of frequency modulation halftoning with sub-dot phase modulation for monochrome image reproduction. These embodiments can, of course, also be used in the context of color printing or, in general, when multiple ink separations are used for printing, for frequency-modulated halftoning of individual color separations. At least one of the color separated images can be halftoned using the method of the present invention.

However, in certain embodiments of the present invention, the concept of sub-dot phase modulation is extended to control the relative positions of the dots in different resolutions and to take into account the overlap characteristics of the dots in different resolution images. This is preferably done in such a way that the resulting halftone pattern in the resulting overprint has a minimum luminance contrast. Since the concept of sub-dot phase modulation allows more precise control of the relative position of the halftone dots of different resolutions in smaller increments, this particular embodiment has a significantly lower level compared to standard techniques. Induces halftoning granularity.

In the following, various steps of the error diffusion method will be described, which use sub-dot phase modulation to optimize the relative positions of the halftone dots of different resolutions when converting an image composed of a plurality of resolution images. By convention, the following assumes that the range of the original color channels is [0.0, 1.0].

The first step consists in determining for each position (i, j) of the image an intermediate luminance channel variable L (i, j) equal to the sum of the N original colorant pixel values at that position.

範 囲 The range of this luminance channel is [0.0, N].

The next step consists of performing multi-level frequency modulation halftoning on this luminance channel. The number of levels is equal to N. In principle, any kind of multi-level frequency modulation halftoning technique can be used, but it is possible to use error diffusion, for example, in combination with any of the known improvements in existing techniques mentioned earlier in this document. preferable. The output of the multi-level error diffusion system is such that each pixel is set {0, 1, 2,. . . It consists of a quantized luminance channel L '(i, j) having a value of N-1｝. The quantized luminance channel represents the optimal luminance distribution of a composite overprint of different colorant channels.

In the third step, the vector error diffusion is performed on the original color channel, using the output of the quantized luminance channel as an additional constraint of the quantizing step. Vector error again preferably employs any of the known improvements in the existing art referred to earlier in this document.

The quantization step of the vector error diffusion process using the concept of sub-dot phase modulation will now be described in detail. The quantization step consists of first determining a set of "valid quantization vector choices" and then selecting the most appropriate one.

To obtain a set of valid quantization choices, first multiply all modified colorant values P '(i, j) x by the number of pixels that make up the halftone cluster. This multiplication recognizes the fact that not all the dots have the same size for the reasons described in the previous embodiment and based on such a technique. The numbers thus obtained are ranked from minimum to maximum. This ranking controls the order of subsequent processing.

For example, in the case of four-color printing, the numbers obtained in the previous ordering step are c ₁ , c ₂ , c ₃ ,. . . it is assumed that c is _four.

The first quantization vector option is to not place any halftone dots of the colorant. According to the method presented in one of a flow diagram illustrating the conventions and the aforementioned embodiments, it is quantized with respect to the four individual ink alternatives _{Q2 (c 1), Q2 (} c 2), Q2 (c 3), And Q ₂ (c ₄ ). The sum of these four values is called QS ₁ and represents the luminance of the first quantization vector choice.

The next possible quantization vector option consists of placing a single colorant halftone dot. When placing the one dot, the first candidate is c _1. This is because this channel has the lowest value and "needs most to darken" depending on the placement of the dots. According to the method presented in one of a flow diagram illustrating the conventions and the aforementioned embodiments, this quantization alternatives _Q1 (c ₁₎ for the four individual inks, Q2 (c 2), Q2 (c 3), and Q2 (C ₄ ) is derived. The sum of these four values is called QS _2, represents the luminance of the second quantization vector alternative.

A third possible quantization vector option consists of placing two colorant halftone dots. Two of the best candidate is _{c 1} and c2. This is because these two channels have the lowest values and are "most needed to darken" by the placement of the dots. According to convention and the method presented in any of the flowcharts describing the previous embodiments, this is the case with the quantization options Q1 (c ₁ ), Q1 (c ₂ ), Q2 (c ₃ ), and Q2 for the four individual inks. (C ₄ ) is derived. The sum of these four values is called QS _3, represents the luminance of the third quantization vector alternative.

In this way, we continue until the five possible quantization vectors represented by their luminance values QS ₁ , QS ₂ , QS ₃ , QS ₄ and QS ₅ have been calculated.

At this stage, it is ready to actually select the quantization vector. The value QS1 _[1-N] that most closely approximates the quantized luminance value L '(i, j) is the selected quantized vector. By using this procedure, it is achieved that the luminance change resulting from the vector quantization follows the distribution L ′ (i, j) resulting from the frequency modulation process on the luminance channel of the original image.

Now that the vector error diffusion quantization step has made it possible to determine which colorant channel should receive the dot at position (i, j) and which channel is not, And work on the actual error diffusion, which means calculating the errors and diffusing them to at least one unprocessed pixel in the corresponding colorant plane. This is performed according to the method disclosed in the previous embodiment. It is not necessary to calculate the luminance values corresponding to all possible quantization vector options. For example, if it is found that QS ₁ <L ′ (i, j) <QS ₂ , there is no need to calculate QS ₃ , QS ₄ , and QS ₅ . This is because the calculation of these additional luminance values does not affect the final result of the selection of the quantization vector option. This allows for improved performance of the described algorithm.

手 順 The above procedure describes in detail the case of four colorants. However, it is easy to extend for N inks.

Many other variations and enhancements of the embodiments described above are possible. For example, use a diffusion weight that varies as a function of colorant value or position in colorant vector space. Other variations include the use of colorants that have essentially the same hue but different densities, or special colorants for the purpose of extending the printable color gamut. As a further variation, the method described above can be used for only a subset of the colorants, or error diffusion can be performed at different resolutions for different colorants.

Example of an Image Processed According to the Above Embodiment FIG. 23 shows an image resulting from the state of the art halftoning method. The output image was obtained by using Floyd and Steinberg error diffusion in combination with staggered scanning and weighted 50% perturbation, followed by 2 × 2 pixel replication. FIG. 24 shows a halftoned image according to the present invention using halftone dots of 2 × 2 pixels. As can be seen, the images obtained using the method according to the invention show fine details and low graininess, while there are few artifacts such as worm effects and patterning near reasonable gradation values. Dots have a lower tendency to join, and the distance between the dots is more precisely controlled. FIG. 25 shows an example using sub-dot phase modulation and error diffusion with variable dot size.

Having described preferred embodiments of the invention in detail, those skilled in the art will now appreciate that many modifications can be made thereto without departing from the scope of the invention as set forth in the claims. There will be.

2 shows the degradation obtained by AM screening. It shows the same degradation as FIG. 1 obtained by FM screening. 1 shows a basic error diffusion method. 4 shows an alternative implementation of the error diffusion scheme. 9 shows frequency modulated halftoning of a 5/9 color tone without pixel duplication. Figure 5 shows frequency modulation halftoning of a value 5/9 tone with 2x2 pixel duplication. 4 shows the results of a method for obtaining clustered halftone dots with improved printing properties using pixel replication. FIG. 8 shows the representation of 128/255 tone values obtained by standard Floyd and Steinberg error diffusion. 19 shows 192/255 tone values obtained by Floyd and Steinberg error diffusion. Shown are 128/255 tone values obtained by Floyd and Steinberg error diffusion using staggered scanning and threshold 50% noise. FIG. 4 shows an example of a possible cyan decomposition with a value of 128/255 obtained by Floyd and Steinberg error diffusion. FIG. 4 shows an example of a possible magenta decomposition with a value of 128/255 obtained by standard Floyd and Steinberg error diffusion. FIG. 13 shows an overprint representation of the disassembly of FIGS. 11 and 12 in a substantially perfect registration state. FIG. 13 shows an overprint representation of the separations of FIGS. 11 and 12 with the two separations being slightly misregistered. The result of 2 × 2 pixel duplication performed following frequency modulation halftoning of a tone having the value 5/9. The result of 2 × 2 pixel duplication performed after frequency modulation halftoning of a value 5/9 tone in which a random element is introduced in order to eliminate an artifact due to a phase correlation halftone dot position is shown. Figure 5 shows frequency modulation halftoning of a 5/9 tone with 2x2 pixel duplication and sub-dot phase modulation. 5 is a flowchart illustrating an alternative interpretation of the standard error diffusion quantization process. 4 is a flowchart describing a quantization process for error diffusion with sub-dot phase modulation. An example of quantization by error diffusion with sub-dot phase modulation is shown. 4 shows an error diffusion method using sub-dot phase modulation and gradation-dependent variable halftone dot size. FIG. 5 illustrates an error diffusion scheme using sub-dot phase modulation and a dot size that depends on both the tone value and the local contrast. FIG. 4 illustrates rendering of an image with frequency modulation halftoning using state-of-the-art error diffusion. 4 shows the result of error diffusion using sub-dot phase modulation. 9 shows the results of sub-dot phase modulation and frequency modulation halftoning using variable size halftone dots.

Claims (4)

A method for frequency modulation halftoning using halftone dots, wherein at least one halftone dot is comprised of a cluster of adjacent pixels arranged on a pixel grid, the method comprising: A method characterized in that it can be placed at any position. 方法 The method of claim 1, wherein the frequency modulation halftoning method is based on an error diffusion algorithm. The method of claim 2, wherein a quantization set is determined for at least one pixel location, and a cluster of at least two pixels is set for at least one quantization value. A transformed pixel value based on the input pixel value and a transformed pixel value consisting of available quantized values, each corresponding to a combination of available output pixel values of a cluster of pixels resulting in a change in density value of the output image. Determining a quantization set;
Selecting a quantization value from the quantization set based on the modified pixel value;
Calculating an error value based on the modified pixel value and the selected quantization value;
Deforming at least one pixel by adding a fraction of the calculated error,
Characterized in that a change in the density value of an area of the output image corresponding to two or more pixels is considered.
3. The method according to claim 2, for converting an image consisting of input pixels into an output image.

Images